README.md

---
pretty_name: RescueBench
tags:
  - embodied-ai
  - robotics
  - search-and-rescue
  - visual-navigation
  - spatial-reasoning
  - unreal-engine
  - benchmark
---

# RescueBench

**RescueBench** is a mission-level benchmark for evaluating embodied agents in simulated search-and-rescue (SAR) environments. Given multimodal clues and egocentric observations, an agent must autonomously explore an unfamiliar environment, locate and approach the target, return to the rescue area using spatial memory, and complete the final handoff.

Each mission is decomposed into four sequential stages:

1. **Explore and locate** the target.
2. **Approach and retrieve** the target.
3. **Return** to the rescue area using observation history and spatial memory.
4. **Complete the handoff** and finish the mission.

This repository hosts the RescueBench evaluation data and representative demonstrations across different environments, agents, and difficulty settings.

## Human-Control Demonstrations

<table>
  <tr>
    <td align="center">
      <img src="./assets/human_forglar_l4_e5.gif" width="260"><br>
      <b>HongKongStreet / Human / L4</b>
    </td>
    <td align="center">
      <img src="./assets/human_tokyo_l3_e34.gif" width="260"><br>
      <b>Tokyo / Human / L3</b>
    </td>
    <td align="center">
      <img src="./assets/human_downtownwest_l2_e44.gif" width="260"><br>
      <b>DowntownWest / Human / L2</b>
    </td>
  </tr>
</table>

These GIFs are representative examples illustrating the human-control interface and complete mission workflow; they are not the full set of participant trajectories.

## Benchmark Demonstrations

<table>
  <tr>
    <td align="center">
      <img src="./assets/Rocket2_L0_p24.gif" width="260"><br>
      <b>FlexibleRoom / ROCKET</b>
    </td>
    <td align="center">
      <img src="./assets/rocket_l0.gif" width="260"><br>
      <b>FlexibleRoom / ROCKET</b>
    </td>
    <td align="center">
      <img src="./assets/yolo_lo_p9.gif" width="260"><br>
      <b>FlexibleRoom / YOLO Planner</b>
    </td>
  </tr>
  <tr>
    <td align="center">
      <img src="./assets/uni_navid_l1_p33.gif" width="260"><br>
      <b>DowntownWest / Uni-NaVid</b>
    </td>
    <td align="center">
      <img src="./assets/vint_forlar_l2p2_8s.gif" width="260"><br>
      <b>Forglar Map / ViNT</b>
    </td>
    <td align="center">
      <img src="./assets/uni_forglar_p7.gif" width="260"><br>
      <b>Forglar Map / Uni-NaVid</b>
    </td>
  </tr>
  <tr>
    <td align="center">
      <img src="./assets/uni_tokyo.gif" width="260"><br>
      <b>Tokyo / Uni-NaVid</b>
    </td>
    <td align="center">
      <img src="./assets/uni_HK_11.gif" width="260"><br>
      <b>HongKongStreet / Uni-NaVid</b>
    </td>
    <td align="center">
      <img src="./assets/Rocket_l1_p10.gif" width="260"><br>
      <b>SuburbNeighborhood Day / ROCKET</b>
    </td>
  </tr>
</table>

## Heterogeneous Multi-Agent Cooperation

<table>
  <tr>
    <td align="center">
      <img src="./assets/third_person.gif" width="260"><br>
      <b>Third-person view</b>
    </td>
    <td align="center">
      <img src="./assets/SPF_p8_first2s.gif" width="260"><br>
      <b>UAV view</b>
    </td>
    <td align="center">
      <img src="./assets/p8_small.gif" width="260"><br>
      <b>Ground-agent view</b>
    </td>
  </tr>
</table>